v2.4 · Now with streaming tool calls

Build with state-of-the-art AI models, in minutes.

One unified API for text generation, embeddings, and tool calling. Ship production-ready AI features with predictable pricing, low latency, and 99.95% uptime.

Start with Quickstart Explore the API

POST /v1/chat/completions

200 OK · 182 ms

Core documentation

Pick your path — everything you need from first request to production.

Quickstart

Make your first AI API request in under five minutes — install the SDK, add a key, and run your first completion.

Get started →

API Reference

Complete reference for every endpoint — request parameters, response schemas, authentication, and error handling.

Browse endpoints →

Developer Guides

Task-oriented guides covering text generation, streaming, tool calling, and production best practices.

Read the guides →

120M+

API calls served daily

182 ms

Median first-token latency

99.95%

Uptime, last 12 months

Getting started

Quickstart

A step-by-step guide to making your first AI API request — SDK installation, API key setup, and a working sample in three short steps.

Install the SDK

Install the official Vertex SDK for your language. The SDK handles authentication, retries, and streaming for you.

Terminal

# Node.js 18+
npm install @vertex-ai/sdk

Set up your API key

Create a key in your dashboard, then export it as an environment variable. Never hard-code keys in source control.

Shell

export VERTEX_API_KEY=sk-••••••••••

Free-tier keys are rate-limited to 60 requests/min. Upgrade for production workloads.

Make your first request

Run this snippet. You'll get a completion back in a single synchronous call — no configuration required.

first-request.js

import Vertex from "@vertex-ai/sdk";

const client = new Vertex();

const res = await client.chat.completions.create({
  model: "vertex-large-2",
  messages: [{ role: "user", content: "Explain embeddings in one sentence." }],
});

console.log(res.choices[0].message.content);

Expected output

"Embeddings are dense vector representations that let a model measure how similar two pieces of text are."

Next steps

Now that your first request works, go deeper with the full reference or the task-oriented guides.

View API Reference Browse guides

POST

/v1/chat/completions

Generate a model response for a conversation. Send a list of messages and receive the assistant's reply, with optional streaming and tool calling.

Request

Authenticate with a bearer token in the Authorization header. Keys are scoped per project.

model — required. One of vertex-large-2, vertex-fast-2
messages — required. Conversation so far
stream — optional, default false
tools — optional function definitions

cURL

Parameters

Name	Type	Default	Description
model	string	—	Model ID to use for the completion.
messages	array	—	Messages so far, oldest first. Max 128 turns.
temperature	number	1.0	Sampling temperature, 0–2. Lower is more deterministic.
max_tokens	integer	4096	Upper bound on generated tokens.
stream	boolean	false	Stream tokens as server-sent events.

Response

Returns a completion object containing the assistant message, token usage, and a stable finish reason.

200 OK · application/json

{
  "id": "cmpl_8f2ka1",
  "object": "chat.completion",
  "model": "vertex-large-2",
  "choices": [{
    "index": 0,
    "message": { "role": "assistant", "content": "Hello! How can I help?" },
    "finish_reason": "stop"
  }],
  "usage": { "prompt_tokens": 12, "completion_tokens": 9 }
}

Errors

Errors use conventional HTTP status codes and always return a machine-readable JSON body.

Code	Message	How to fix
400	Invalid request: `model` is required	Check the request body against the schema.
401	Invalid or missing API key	Verify the `Authorization` header.
429	Rate limit exceeded	Back off exponentially; check `Retry-After`.
500	Internal server error	Retry with jitter; contact support if persistent.